Pit: Pre-trade Integrity Toolkit

Pit is a workspace for embeddable pre-trade risk components.

The project is built around a simple idea: before an order reaches a venue, it should pass through a deterministic risk pipeline that can reject, reserve state, and later absorb post-trade outcomes back into the same control system.

How The Flow Works

1. The engine is configured once during platform initialization.
2. Each order first goes through lightweight start-stage checks. This stage is meant for fast decisions and for controls that must observe every request, even when the order is rejected early.
3. If the order passes the start stage, the caller receives a deferred request object. That object represents the heavier pre-trade stage, but the heavy checks have not run yet.
4. When the deferred request is executed, the engine runs the main-stage risk policies and collects all registered mutations. If the stage fails, the collected state is rolled back. If it succeeds, the caller receives a reservation.
5. The reservation must then be finalized explicitly. Commit keeps the reserved state. Rollback cancels it. Dropping the reservation without finalization rolls it back automatically.
6. After execution, post-trade reports are fed back into the engine so that policies depending on realized outcomes can update their internal state.

Current Scope

The current implementation focuses on the pre-trade pipeline and a set of foundational controls:

• P&L kill switch
• sliding-window rate limit
• per-settlement order size limits

The engine is intentionally in-memory and deterministic. It is designed to be embedded into a larger trading system rather than replace one.

Where To Start

The openpit crate README if you want to start with the Rust interface and a runnable example.

The Python bindings README if you want to work with Pit from Python via the openpit package.

Conceptual pages and longer architecture notes wiki..

Local Build And Test

Prerequisites

• Rust toolchain
• Python >=3.9
• maturin for Python bindings build
• pytest for Python tests

python3 -m venv .venv
source .venv/bin/activate
pip install -r ./requirements.txt

Build

SDK

With Just:

just build

Manual:

cargo build --workspace

Python

With Just:

just python-develop
just python-develop-release

Manual:

maturin develop --manifest-path bindings/python/Cargo.toml
maturin develop --release --manifest-path bindings/python/Cargo.toml

The recommended Python test flow is to run maturin develop before pytest. This runs against the current checkout (including a dirty worktree). If sources did not change, Cargo can reuse cached artifacts and avoid a full rebuild.

Tests

With Just:

# All tests:
just test-all

# Rust:
just test-rust

# Python:
just test-python
just test-python-unit
just test-python-integration

Manual:

# All tests:
cargo test --workspace
maturin develop --manifest-path bindings/python/Cargo.toml
python -m pytest bindings/python/tests

# Rust:
cargo test --workspace

# Python
maturin develop --manifest-path bindings/python/Cargo.toml
python -m pytest bindings/python/tests
python -m pytest bindings/python/tests/unit
python -m pytest bindings/python/tests/integration